New Research Shows Limits of Risk Adjustment in Protecting Traditional Medicare under Premium Support

Proposals to convert Medicare to a “premium support” system would replace its guarantee of health coverage with a flat payment, or voucher, that beneficiaries would use to purchase either private health insurance or, in some versions, a form of traditional fee-for-service Medicare.  Proponents of premium support argue that traditional Medicare would remain a viable option for beneficiaries under proposals that retain it.  That, however, is unlikely to be the case over the long run as findings from new analyses from the Medicare Payment Advisory Commission (MedPAC) and researchers at the Department of Health and Human Services (HHS) indicate.

Under premium support, traditional Medicare would tend to attract less healthy beneficiaries who cost more to cover, while private plans would tend to attract healthier-than-average beneficiaries, just as private Medicare Advantage plans always have.  As Medicare Advantage plans can do today, private plans could tailor their benefit packages — and use marketing and other tools — to appeal to healthier beneficiaries and deter sicker ones.

Premium support proponents contend that this problem of adverse selection (the separation of healthier and less-healthy people into different insurance arrangements) can be addressed through a risk adjustment system, which would raise or lower payments to private plans and traditional Medicare based on their enrollees’ health status.  But risk adjustment is imperfect and captures only part, rather than all, of the cost variation resulting from differences in beneficiaries’ health.  As a result, premiums for traditional Medicare would inevitably rise because traditional Medicare would receive only partial compensation for its beneficiaries with above-average costs.  The higher premiums would then encourage greater numbers of healthier enrollees in traditional Medicare to switch to private plans, which in turn could set off a cycle of further premium increases for traditional Medicare and reduced enrollment.  Traditional Medicare would thus risk unraveling over time.[1]

While research from Harvard health economist Joseph Newhouse and others finds that the accuracy of risk adjustment in Medicare has improved considerably over time,[2] new analyses examining Medicare risk adjustment, from MedPAC (Congress’ official advisory body on Medicare payment policy) and researchers at HHS, highlight risk adjustment’s continued limitations in fully addressing adverse selection concerns:

  • A June MedPAC report noted that the risk adjustment system now used to pay Medicare Advantage plans is considerably more accurate than in the past but still underpredicts the cost of covering people with higher-than-average costs and overpredicts the cost of covering people with lower-than-average costs.  As a result, Medicare Advantage payments only partially compensate plans with high-cost enrollees and overcompensate plans with low-cost enrollees. 

    MedPAC estimated that Medicare Advantage risk adjustment in 2010 and 2011 predicted only 71 percent of actual costs for individuals above the 99th percentile (that is, people whose costs exceeded those of 99 percent of beneficiaries) and 82 percent of actual costs for those in the 95th to 99th percentiles.  Conversely, risk adjustment predicted costs that were 162 percent of actual costs for low-cost individuals (those in the 20th percentile).[3]
  • Any risk adjustment system requires an accurate measure of each enrollee’s health status, usually expressed as a “risk score” based on diagnoses.  But risk scores are vulnerable to the phenomenon of upcoding (also known as increased coding intensity or coding creep), where the risk scores that private plans submit increase over time without actual changes in enrollees’ health.  This distorts the results of risk adjustment, further reducing its effectiveness.  Under Medicare Advantage, upcoding has been a persistent problem.  Risk scores for Medicare Advantage plans have risen over time even though there appears to have been no corresponding change in the health status of Medicare Advantage beneficiaries, as both the Government Accountability Office and the Congressional Budget Office have previously found.[4] 

    Researchers from HHS’s Agency for Healthcare Research and Quality and the Office of the Assistant Secretary for Planning and Evaluation recently confirmed that risk scores in Medicare Advantage plans have increased faster than those in traditional Medicare and appear largely to reflect changes in diagnostic coding rather than in health status.  The average risk score rose 2.2 percentage points faster per year in Medicare Advantage than in traditional Medicare between 2004 and 2013, assuming the 2004 methodology for how risk scores were calculated was in effect over the entire period.  (The average risk score rose 1.6 percentage points faster per year in Medicare Advantage than in traditional Medicare assuming the system used in 2014 was in place throughout the ten-year period.)  There were no corresponding changes in mortality rates, self-reported health status, or demographic characteristics like age.[5] 

Risk adjustment will likely always remain imperfect to a significant degree, and as a result, risk adjustment will continue to undercompensate plans with costlier-than-average enrollees and overcompensate plans whose enrollees have lower-than-average costs.  No system can fully account for the wide variation (and skewed distribution) in health spending among individuals, as well as the hard-to-predict changes in individuals’ health status and spending from year to year.[6]  Moreover, as MedPAC points out, if plans collect more detailed information about the health of their enrollees over the course of enrollment than the information made available to the entity that administers risk adjustment, the plans can then use that information to exploit limitations in the risk adjustment process to their advantage.[7]  That includes using benefit design and marketing based on this information in ways that can attract or retain enrollees who are the most profitable under the risk adjustment system while discouraging enrollment by others. 

In addition, under a premium support system, it would likely be very difficult politically to enact, implement, and rigorously enforce an accurate risk adjustment system over time.  The kind of active regulatory oversight, reporting and other requirements, and continuous efforts to improve risk adjustment that would be needed[8] are inconsistent with the less regulated approach favored by many premium support proponents.  It is instructive that a premium support proposal designed by Henry Aaron and Robert Reischauer in 1995 included numerous features to limit the risk of adverse selection, including, for example, a requirement that all plan options offer standardized benefit packages.  Those Aaron-Reischauer provisions, however, are missing from recent premium support proposals.  As Aaron stated in 2012, “[c]urrent plans do not contain these elements, and even if they did, there is little prospect in the contemporary political environment of enacting or sustaining them.”  That’s why Aaron has concluded that premium support proposals in their current form are not a sound idea.[9] 

Finally, the consequences of inadequate risk adjustment would be much greater under premium support than in Medicare today.[10]  If risk adjustment does not work well in Medicare Advantage, for example, it results in higher payments to such plans than are warranted, a worsening of Medicare’s finances, and somewhat higher Part B premiums for beneficiaries.  Under premium support, however, traditional Medicare itself would be fundamentally threatened.  As an Urban Institute study noted, without a sufficiently accurate risk adjustment mechanism, “plans (such as traditional Medicare) that attracted disabled or less healthy enrollees would be at a competitive disadvantage and subject to financial losses, and beneficiaries who might be expected to require more care would face higher premiums and severe access problems.”[11]  Over time, traditional Medicare could become increasingly unsustainable.

In short, risk adjustment is no panacea.  Under premium support, adverse selection would likely remain a substantial risk that could threaten traditional Medicare’s long-term viability.

End notes:

[1] See Paul N. Van de Water, “Medicare in Ryan’s 2015 Budget,” Center on Budget and Policy Priorities, April 8, 2014, https://www.cbpp.org/cms/?fa=view&id=4121.  See also Edwin Park, “Ensuring Effective Risk Adjustment,” Center on Budget and Policy Priorities, May 18, 2011, https://www.cbpp.org/cms/index.cfm?fa=view&id=3497.

[2] Joseph Newhouse and Thomas McGuire, “How Successful Is Medicare Advantage?,” The Milbank Quarterly, Vol. 92: 2, 2014 and J. Michael McWilliams, John Hsu, and Joseph Newhouse, “New Risk-Adjustment System Was Associated with Reduced Favorable Selection in Medicare Advantage,” Health Affairs, December 2012.

[3] Medicare Payment Advisory Commission, “Report to the Congress: Medicare and the Health Care Delivery System,” June 2014, chapter 2, http://medpac.gov/chapters/Jun14_Ch02.pdf.

[4] See Government Accountability Office, “Medicare Advantage: CMS Should Improve the Accuracy of Risk Score Adjustments for Diagnostic Coding Practices,” January 26, 2012, http://gao.gov/assets/590/587637.pdf, and Congressional Budget Office, “Designing a Premium Support System for Medicare,” December 2006, http://www.cbo.gov/sites/default/files/cbofiles/ftpdocs/76xx/doc7697/12-08-medicare.pdf.  See also January Angeles and Edwin Park, “‘Upcoding’ Problem Exacerbates Overpayments to Medicare Advantage Plans,” Center on Budget and Policy Priorities, September 14, 2009, https://www.cbpp.org/cms/index.cfm?fa=view&id=2712

[5] Richard Kronick and W. Pete Welch, “Measuring Coding Intensity in the Medicare Advantage Program,” Medicare and Medicaid Research Review, Vol. 4 No. 2, 2014, http://www.cms.gov/mmrr/Downloads/MMRR2014_004_02_sa06.pdf.  Note that the Newhouse-McGuire analysis did not appear to take into account the effects of upcoding trends on the effectiveness of Medicare Advantage risk adjustment.

[6] See, for example, Park, op cit.

[7] Medicare Payment Advisory Commission, op cit.

[8] The Centers for Medicare and Medicaid Services (CMS) now periodically updates its risk adjustment models for Medicare Advantage and Medicare Part D, institutes annual adjustments to take into account coding intensity trends, and conducts retrospective audits of how private plans assign diagnoses (known as Risk Adjustment Data Validation audits), which can help ensure that enrollee diagnoses are consistent with medical records and other documentation.  Medicare Advantage insurers have strongly opposed these efforts and, in some cases, have succeeded in pressuring CMS to drop initiatives to improve risk adjustment accuracy.  To try to combat adverse selection adequately under a premium support system, however, all of these elements would be needed along with additional regulations, such as those Aaron and Reischauer outlined.

[9] Henry Aaron and Austin Frakt, “Why Now Is Not the Time for Premium Support,” New England Journal of Medicine, 366:877-879, March 8, 2012, http://www.nejm.org/doi/full/10.1056/NEJMp1200448.

[10] Paul N. Van de Water, “Converting Medicare to Premium Support Would Likely Lead to Two-Tier Health Care System,” Center on Budget and Policy Priorities, September 26, 2011, https://www.cbpp.org/cms/?fa=view&id=3589.

[11] Stuart Guterman, “Risk Adjustment in a Competitive Medicare System with Premium Support,” in Marilyn Moon (ed.), Competition with Constraints: Challenges Facing Medicare Reform (Washington: Urban Institute, 2000).